AI classroom ban left students last, Schrepel finds

The experiment placed students into three different relationships with ChatGPT: complete denial, uncritical access, and structured training. The results showed that the training gave a large initial boost—especially on the take-home exam—but that boost disappeared a year later as students became more familiar with AI tools in daily life. Meanwhile, both the no-AI group and the unguided group produced unexpected outcomes: the former ran out of ideas within minutes, while the latter avoided repeating their classroom mistakes on exams and even edged out the AI-free group. The author, Schrepel, had expected the opposite but now says he was wrong to assume AI only helps with formal guidance.

These findings sit alongside other research. A UC Berkeley study of over 500,000 grades found that A grades in writing- and programming-heavy courses jumped 13 percentage points after ChatGPT launched, while proctored exams showed no comparable shift. A study from central China tracking more than 26,000 K-12 students found AI boosted homework grades but exam performance dropped by up to 24 percent. Together, these suggest that the setting—unsupervised versus proctored—may matter as much as whether AI is allowed at all.

Schrepel's conclusions—resist blanket bans, invest in faculty skills, and rethink the master's thesis—will be tested as universities like UC Berkeley Law move in the opposite direction. The outcome may hinge on whether institutions prioritize building foundational thinking before AI use, or accept that students are already using these tools and need to learn to use them well. Schrepel himself acknowledges the study's small sample and the difficulty of tracking actual AI use, so the debate is likely to continue.